Motor skill training and transcranial direct current stimulation (tDCS) have separately been shown to alter cortical excitability and enhance motor function in humans. Their combination is appealing for augmenting motor recovery in stroke patients, and this is an area presently under heavy investigation globally. We had previously proposed that the success of transcranial direct current stimulation (tDCS) as a complementary therapeutic tool in physical neurorehabilitation may be contingent upon a more precise knowledge of the circumstances under which the two techniques interact, and the subsequent behavioral outcome. To test this we applied tDCS with conventional parameters (polarity, location, intensity and duration), but manipulated the timing of application in relation to a simple robotic motor practice task, in affected muscles of hemiplegic patients. We showed a profound effect of timing. We showed that performance improvement over one session of practice was differentially affected by tDCS depending on the timing of application. 20 mins of robotic visuomotor training only, led to a preferential increase in movement speed without a loss of accuracy. tDCS applied at any of the timings obliterated this effect, yet when tDCS was applied before training, the effect was substituted for an improvement in smoothness. Movement smoothness is associated with more advanced stages of motor learning and has high correlation with functional clinical scales. We showed for the first time that the timing of tDCS application has functional significance, that tDCS applied prior to training can be beneficial for voluntary behavior, and that tDCS effects may not simply be additive to training effects, but may change the nature of the training effect. We have separately reported in a randomized-controlled clinical trial, that upperlimb robotic training alone over 12 weeks can improve clinical function of chronic stroke patients. Based on our results with tDCS and robotic training, we hypothesize that the same repeated sessions of robotic training, but preceded by tDCS, would lead to a sustained and functional change greater than robotic training alone. In the proposed training study, 66 stable chronic stroke patients will be randomized to receive sham or real tDCS prior to robotic upperlimb training, and will be assessed at 3 levels of change: (i) clinical, (ii) kinematic, and (iii) neurophysiologic. We will determine if clinical function can be improved and sustained with tDCS-robotic training, the precise kinematic aspects of movement, and cortical physiology changes that underlie functional improvements.